Who to hire for Krux on Salesforce

A data management platform and a customer data platform solve neighbouring problems in quite different ways, which is the central fact for anyone trying to hire after Krux. The DMP existed to collect anonymous web and mobile signals, sort them into audiences and push those audiences to advertising and personalisation destinations. Data 360, formerly Data Cloud, is built around known and unified customer profiles inside your own Salesforce environment. A partner who treats the second as a drop-in replacement for the first will disappoint you, and the disappointment will arrive late and expensively.

#The facts a partner must get right

LiveRamp's customer notice put the end date in one line: "Salesforce has officially announced that the End of Life date for Salesforce Audience Studio DMP (formerly known as Krux) is February 1, 2024." The same notice adds that "Salesforce had previously announced an upcoming End of Life for Audience Studio DMP in July of 2021." and lists the practical consequences: existing first-party and marketplace distributions were turned off, and existing destination accounts could no longer be viewed or activated. You can read it in full on LiveRamp's documentation. Several commentators observed that Salesforce steered customers towards its customer data platform, and that is consistent with the product direction, but no partner should tell you that the move is automatic or that the old audiences carry across.

The first hiring test is therefore cheap. Ask what happened to your Krux audiences on 1 February 2024 and what the candidate would need in order to rebuild them. A good answer starts with questions about where the source data lives, not with a Data 360 demonstration.

#Why this is a different skill from a Krux project

Krux practitioners think in tags, pixels, segments and syndicated data. Data 360 practitioners think in data streams, mappings to a data model, identity resolution rulesets, unified profiles, calculated insights, segments and activation targets. Some skills overlap, and plenty do not.

The last row is the one that separates honest partners from sellers. Third-party marketplace audiences had no like-for-like successor, and a partner should say what you will do without them.

#Certifications and experience to look for

Start with Salesforce Certified Data 360 Consultant, the credential formerly named Data Cloud Consultant, which tests ingestion, harmonisation, identity resolution, segmentation and activation. Check that the person holds the current exam rather than an older one, as the exam guide was revised with the rename. Add a Marketing Cloud credential only if your activation runs through Marketing Cloud. Add Platform Developer I if custom code will touch ingestion or activation. Beyond that, ask for data engineering experience in the warehouse or lake that feeds Data 360, because most delivery problems begin upstream.

Experience matters more than papers. Ask the candidate to describe a time identity resolution merged people who should not have been merged, and how they found it. Ask how they managed consent flags from a preference centre into profiles. A person who has run this in production has a story.

#A realistic first engagement

Start with an audience inventory rather than a platform build. The deliverable is a table of every audience you used on Krux, who used it, where it was sent, what data created it and whether the data still exists. Then a source readiness review: for each source, is there a connector, is there consent, and how clean are the identifiers? Finally, one narrow pilot: choose a single valuable audience, build it end to end in Data 360, activate it to one destination, and measure whether it behaves as the old one did.

Scope the pilot so it fails safely. Use a sandbox where possible, limit activation volume, and agree the acceptance test before building. If the pilot works, the larger programme has evidence behind it. If it does not, you have learned that cheaply. For how cost is built up, see Krux implementation cost and the Krux licence page.

#Where audience projects really fail

Three failures account for most of the pain after a DMP is retired, and a good partner will raise all of them before you do.

The first is consent. A DMP lived largely on anonymous, tag-based signals whose consent was managed at the website. A customer data platform holds profiles that can be linked to a person, which raises the bar. Someone has to decide which consent flags exist, where they come from, how a withdrawal travels to every destination, and what happens to a profile when a person asks to be forgotten. Your privacy officer should be in the room before the first stream is configured, not after the first segment is activated.

The second is identity. Matching rules decide whether two records are one person. Loose rules merge strangers and send the wrong offer to the wrong household. Tight rules leave one customer split in three, and every audience double counts. A partner should propose a starting ruleset, a way to sample and inspect merges, and a rule for reviewing the results each month. The word to listen for is evidence. Anyone can describe fuzzy matching, and few can show you the sample they reviewed.

The third is activation. Destinations have their own matching requirements, minimum audience sizes and delay between sending and appearing. An audience that looks healthy in Data 360 may show a fraction of its size in an advertising platform, and a partner who has done this before will already have an explanation and a monitoring routine.

#Who should be on the team

A single consultant rarely covers it. Look for a lead who can talk to marketing about outcomes, a data engineer who can work with your warehouse or lake and the Data 360 connectors, a person who understands consent and privacy design, and someone who knows the destinations you activate to. In a small engagement these may be two people, but the four skills must exist. Ask the firm to name them, and ask which of them you will actually see. Where a partner proposes to staff the data engineering with juniors and the lead with a seller, you can expect the first month to be slow and the second month to be about rework.

Also ask about Salesforce-side knowledge you may need in parallel. If activation runs through Marketing Cloud Engagement, someone must manage data extensions and journeys. If profiles feed Sales or Service, someone must decide which unified fields appear on which record. These are separate decisions from the audience build and often belong to different teams in your organisation.

#Red flags specific to this situation

"Identity resolution gives you a golden record" is the most common overclaim. Identity resolution in Data 360 produces unified profiles by rules, and those rules can be wrong. A partner who promises a single perfect view has not run it on messy data. "We will migrate your Krux audiences" is another, since the product ended and there is nothing to migrate except your own exports and definitions. A proposal with no consent section is a third. So is a team that plans activation before checking the destination's matching requirements, and one that sets no limits on credits or volumes, since Data 360 consumption is metered.

#Ten questions for the interview

1. Which of my Krux audiences can be rebuilt from first-party data, and which cannot?
2. What identity rules would you start with, and how would you measure false merges?
3. How do consent and opt-out flags travel from source to activation?
4. Which sources would you ingest first, and why?
5. How would you handle anonymous visitors who later log in?
6. What do you do when a destination matches fewer people than expected?
7. How do you estimate and monitor Data 360 consumption for a pilot?
8. Describe a segment you built that behaved differently from the business expectation.
9. How do you document data lineage so a successor can follow it?
10. What would you tell me not to build in Data 360?

Question 10 is the most revealing. A partner who can name three things Data 360 should not be used for has used it.

#What to do with Krux-era data you still hold

Check what survived. Some organisations kept exports of audience definitions, tag inventories, data dictionaries or reports from the old platform. Those documents are more useful than they look, because they record which signals your marketing team believed predicted value. A partner can use them to prioritise what to rebuild and to find audiences that nobody has used for years and can be dropped. If you hold nothing, the same discovery has to be done through interviews with the people who requested audiences, and the work takes longer.

Be sceptical of any suggestion that old cookies or device identifiers can be reloaded to recreate previous reach. Browser changes, consent rules and the end of the DMP itself all make that unlikely, and a partner who proposes it should explain the lawful basis.

#Measuring whether the rebuild worked

Agree the yardstick before the work begins. For each rebuilt audience, decide how you will compare it with the old one: size, overlap with the previous membership if you have records, match rate at each destination, and the commercial result of the campaigns that used it. Accept that some differences are improvements, such as fewer stale cookies and better consent records, and some are losses, such as reach that was never first-party in the first place. A partner who sets this out in the proposal, with a review date, is planning for accountability. One who promises parity is selling comfort.

#Checking evidence without relying on logos

Ask for a reference whose project involved identity resolution across at least three sources, and ask the reference how many unresolved or wrongly merged profiles surfaced in the first month. Ask for a data model diagram from a past engagement with names removed. Check credentials on Salesforce's public site. Ask who wrote the consent design, because it is often a gap that no one noticed until legal asked.

#Commercial models and the statement of work

The inventory and readiness review suit fixed fees. The pilot suits a fixed fee with a cap on consumption. A larger rollout suits phases with acceptance criteria and a change process, because new sources keep appearing. The statement of work should name the sources and destinations, the identity rules and who approves them, the consent design, consumption monitoring and who pays when it exceeds plan, the data model, documentation and the handover. Related reading: what happened to Krux, should I migrate from Krux and integrating Krux with Salesforce.

#When you can skip a partner

If your audiences were small, built from first-party data you already hold in Salesforce or Marketing Cloud, and used for one or two destinations, a certified in-house analyst can build them in segmentation tools without a partner. Bring one in for multi-source identity, consent complexity or high activation volumes. This page ranks no agency and recommends no partner.